Insights into COVID-19 Dynamics via the Integration of Subpopulations, Vaccinations, and Bayesian Calibration in a Semi-Enclosed Community Model Framework

From the beginning of the COVID-19 pandemic, universities have experienced unique challenges due to their multifaceted nature as a place of education, residence, and employment. Previous work has used mathematical models to explore varying approaches to combating COVID-19 and tailored these models to local contexts, such as hospitals. However, previous work has not broadly incorporated pharmaceutical mitigation strategies into university models. We address this gap in research by integrating subpopulations, vaccines, and COVID-19 variants into a semi-enclosed population model framework. We further improve the quantification of uncertainty in the model parameters by implementing a formal Bayesian model calibration, fusing real-world data (positive tests and the isolation population) to diagnose and address model shortcomings. We further develop our model to account for sudden decreases in vaccine efficacy and leverage Bayesian changepoint detection to determine the scenarios in which we can successfully detect a sudden decrease in vaccine effectiveness at the population level. Through a Sobol’ global sensitivity analysis, we characterize the sensitivity of modeled infection rates to the uncertain model parameters. We demonstrate increased model accuracy after addressing the model shortcomings highlighted by incorporating real-world data. Lastly, we show the safe operating space for detecting sudden decreases in vaccine efficacy.